Convince me

I talked a little while ago about how to design an experiment, so I guess it’s time to talk about the next step in science: writing a research paper for publication. Depending on the type of study you’ve done (clinical vs lab-based vs theoretical), there’s going to be a ton of differences in the style of write-up you choose. But the first step is always the same: organize your data and make sure it’s telling a complete story. And I don’t mean the “once upon a time” kind.

It’s important to build a narrative with your results, where one discovery leads to the next, with a “hypothesis -> data -> hypothesis -> data… -> conclusions” layout. Otherwise your article is going to be confusing and dead boring. The readers don't care about the chronological order of your experiment because NO ONE likes stream of conciousness writing. Ok, maybe literary types do. But scientists just want to be taken on on interesting, logical trip that is well supported.

But what do I mean by well supported? Well, say you're looking at whether eating paste causes chronic diarrhea. (Hey, I have a little kid who just did this for the first time last week. It's relevant!) First, you'd probably want to simply feed paste to some mice and see what happens. Of course, to do this, you'd need a couple of months and tons of paperwork to get ethical approval for any animal study, but that's fodder for another blog post. Let's assume you got the go-ahead, ran the experiments and are now wanting to write it into a journal article.

So the hypothesis for figure 1 reads something like "daily consumption of paste by mice for 1 week results in chronic diarrhea". You run the study, and poo-reka! Hypothesis confirmed. Sorry, scatalogical humour is a side-effect of doing gastroenterology research. You have to be able to laugh it off when colonoscopy pictures are routinely shown during the lunch seminar.

Anyway, that data alone gets the death knell label of "descriptive". You described something interesting and hopefully new, but it doesn't tell us why or how it happens. You need more data. So you move on to the next figure and hypothesize that the paste causes diarrhea by interacting with the lining of the intestinal tract. This is where it gets a little murkier. Since this can't be directly tested (are YOU going to shrink yourself down to go chat with the mouse's colon? I'm comfortable here on my couch, thanks), you probably decide to test things out in a less complicated model first.

Enter, a test tube full of lab-grown colon cells. This sort of in vitro study (as opposed to in a live animal which is in vivo), is very useful to simplify experiments so you can ask very direct questions. For example, does the paste cause water to be transported across the intestinal cells differently which could lead to diarrhea? This is where a lot of time and effort goes in to research. You have to keep hypothesizing and testing until something turns out to be right. Not only are animals too complex (there's so much more than just colon cells in your colon; immune cells, muscle cells, bacteria, etc), they're also expensive, variable and hard to justify ethically when you're just looking at basic concepts.

So you mess around in the lab until you've clarified what you think is happening mechanistically (the buff older brother of "descriptive"). Then you can go back into the mouse model and see if your assumptions hold true. This is often called "proof of principle" and is absolutely required for higher end journals which have very demanding reviewers. At this point, you really want to look over the figures and make sure you haven't forgotten to test something important that might completely undermine your conclusions: a hole in your story. That's what I mean by well-supported.

Basically, the first step in writing a journal article is to take a look at the data you have and decide whether it would convince a fellow skeptical researcher that what you're proposing actually happens. Because that's essentially what the peer review process is. We're our own harshest critics by design so that hoepfully we can weed out over-stated conclusions that make it look like an in vitro study cured cancer. That's one of my pet peeves.

But on the other hand, because we are constantly covering our scholarly butts with maybes and potentiallys, we can look uncertain of our results. Within the science community this is respected, but in the broader world it can lead to doubt and misgivings toward science. There's a fine balance between these extremes when writing that is hard to hit, but well worth the effort.

So now that you have your experimental data all lined up, you need to write up the paper! The next thing to decide is what style to use, and that depends on what journal you're going to submit it to. Which is an interesting area that I'll talk about next post.